xrk971 Pocket Class A Headamp GB

[xrk971]

Hi Byrtt,
This is really cool. Thank you for providing this to the community - I wonder if perhaps we should start a new thread for this over in Headphone Amps section? So the EQ magic takes a cheap headphone and makes it sound like a nicer cheap headphone? I wonder how much more magic there is when you do this to a DT-880 or HD-600? I have the DT-880 so would like to see that correction curve if you ever have time. I know that manually tweaking 35 PEQ's is not easy! So you deserve a huge thanks!

[BYRTT]

Hi xrk971,
Admit forgot to write sound was based on first impression they sound the same so lets see if that holds in long run and with more different ears on it, there is differences we know of if for example we zoom in on impedance curves it discover DT-880 is smoother than OB-1 so engineering for DT-880 probably focused there and took much more time and knowledge to get it improved in smaller acoustic resonances not taking place into final design, also nominal impedance will make a difference for PCA performance and level of boost at low end for HPS-5000 verse OB-1 is a lot more. HD-650 is then again smother in impedance than DT-880 so we probably in long run and for long time listening get some small improvements, but what do i know about this other that these EQ correction works fantastic for OB-1 plus HPS-5000 : ) will create a DT-880 correction so we get more ears and opinion if I'm right for sound in long run.

[babysupra]

Originally Posted by BYRTT

Traced also average response of dirty cheap Behringer HPS-5000 and created correction to hit Harman curve and must say in corrected stage OB-1 verse HPS-5000 now start sound very much exactly the same, for long listening sessions HPS-5000 wins because they don't clamp so hard against head and also weight much less but in they cheap don't have any nice look or feel of nice build quality.


HPS-5000 correction below, black is Harman target response curve, green is traced average of HPS-5000 raw curves measured over innerfidelity website, blue is correction curve which takes 35x EQ filters or one convolution 32bit wav-file, and purple is green times blue showing fair close coherence to black target.


OB-1 correction below, black is Harman target response curve, brown is traced average of OB-1 raw curves measured over innerfidelity website, orange is correction curve which takes 25x EQ filters or one convolution 32bit wav-file, and green is brown times orange showing fair close coherence to black target.



Can share numbers for these HPS-5000 filters too so if any is interested give a note.

Hi Brett,

could you explain how to create the correction curve for different headphones? I want to try correct my other headphones/earphones if possible

thanks!

[BYRTT]

Sure babysupra, plan is tomorrow make DT-880 correction for xrk971 and thereafter will make guide in MS Publisher how response correction is created using free tools FPGraphTracer/REW/Rephase and try get guide converted and shared as pdf document.

[xrk971]

I will take a guess here based on the graphing of REW. I think Byrtt digitizes the curves from Innerfidelity. Imports the curve for the headphone, then imports the Harman curve. They are loaded into REW where there is an EQ dialog. Then start tweaking PEQ's while watching the fit - manually adjust until the processed curve matches the target. Transfer those PEQ's into Jriver.

[xrk971]

Byrtt beat me to it. Thanks for making the DT880 curve!

[BYRTT]

Originally Posted by xrk971

Byrtt beat me to it. Thanks for making the DT880 curve!

Making me a PCA is also a big thanks.

About guessing its close, difference is EQ correction is done in Rephase because it has nearly unlimited banks of PEQs and more flexiable settings than into REW, for example REW shelving filters have no way to adjust Q, also Rephase can output a convolution file into various formats depending what box is used or one can simply copy numbers from filters set into some player software.


Some can maybe do task in smarter way but its the software i learned to manage and they all free except i support some little donation from time to time because its great tools, more specific info tomorrow in a guide but until that here is headlines:


1. Capture grey curves from Innerfidelity pdf document with MS snipertool saved as png, then in MS paint draw a purple line in middle of those grey curves which should be the average response and save png file.

2. Trace that purple response curve in FPGraphTracer and save as frd file, also Harman curve is traced same way once for all and can be reused for every new set of head phones one like to correct in future.

3. In REW import those two curves and offsets head phones curve relative to Harman and zero dB point so as max of boost and cuts will be kind of symmetric in dB value hovering around zero dB line. Now at "ALL SPL" tab REW can do the math and create a difference curve between Harman curve and actual head phone curve picking math (A over B), and this new difference curve is exported as frd or txt file.

4. In Rephase above is imported and then inverted because smart thing is if this difference curve between Harman target curve or it could be another famous head phone curve say HD-650 with difference to own head phone curve is EQed visual flat in Rephase we have our filter correction numbers or can output a single convolution file. This is the step that takes most time puzzling with sweeping Q/dB/frq numbers up and down until visual curves resonances and deviation from flat start get good as possible looking, think this phase probably needs a bit of patience even one is good at it.

[stellarelephant]

Originally Posted by xrk971

So the EQ magic takes a cheap headphone and makes it sound like a nicer cheap headphone?

EQ can do an awesome job of flattening the amplitude across the usable frequency range of a driver. However, remember that the cause of all those peaks in the frequency response are breakup modes and enclosure resonances. When we smash down the amplitude of the signal at these "ringing" frequencies, we make it so that the music PLUS the unwanted ringing is now equal in amplitude to the rest of the audio spectrum. Those frequencies will still have distortion as part of their makeup.

Is EQing still worthwhile? I think so. But I believe this is why you can't make any old pair of cans sing like a STAX.

----------------------

In other news, my new PCA is a BEAST!

I upgraded pretty much everything for this build, so it is hard to nail down what is responsible for what, but I have a feeling the 64mA bias current is a big part of the "zing" factor. The new caps make a big difference too. As I mentioned, the PPS input caps really open up the treble going into the transistors. Right now I am running my amp with 25V 1000uF Panasonic FR on the output. These are clearly more detailed and neutral than the Nichicon KA I used last go round. No bypass--none needed! This thing has treble for days...

The sound now is much faster and more "technical" than my first build. First one was warm and shined with jazz and acoustic music. This one is agile and transparent, and can handle anything without breaking a sweat.

I just threw one of the most difficult reference tracks I have in my collection at it..."Letting Go" by BT, JES, and Fractal. It has everything from female vocals to bass drops to glitchy stop-and-go synth stabs to lightning speed 128th-note drum machine farts, lol.

Listening to this mess with a pair of ATH-M50 cans, my impressions are:
1) Best bass I've ever heard out of a headphone
2) Detail is insanely good
3) Control! This thing may run hot to the touch, but the presentation is cool as a cucumber. It maintains its composure through even the most complex passages. Crystal clear, powerful, and engaging sound.

[babysupra]

Originally Posted by BYRTT

Sure babysupra, plan is tomorrow make DT-880 correction for xrk971 and thereafter will make guide in MS Publisher how response correction is created using free tools FPGraphTracer/REW/Rephase and try get guide converted and shared as pdf document.

That would be super awesome!

It's going to be nice to have my Monoprice 8323 and Koss KSC75 sounding like HD650 huehuehue

Which makes me wonder if the bass roll off in KSC75 is hardware limitation or due to tuning, because i like me some bass hahaha

[xrk971]

StellarE,
Sounds like you have yourself a killer rig. I have to try these Panasonic FR's and PPS caps.

Listening to this mess with a pair of ATH-M50 cans, my impressions are:
1) Best bass I've ever heard out of a headphone
2) Detail is insanely good
3) Control! This thing may run hot to the touch, but the presentation is cool as a cucumber. It maintains its composure through even the most complex passages. Crystal clear, powerful, and engaging sound.

Doesn't get much better than that.

Thanks for the feedback.

EQ can do an awesome job of flattening the amplitude across the usable frequency range of a driver. However, remember that the cause of all those peaks in the frequency response are breakup modes and enclosure resonances. When we smash down the amplitude of the signal at these "ringing" frequencies, we make it so that the music PLUS the unwanted ringing is now equal in amplitude to the rest of the audio spectrum. Those frequencies will still have distortion as part of their makeup.

A lot also depends on how much headroom is left after the EQ has been applied. Generally, I don't EQ dips, except broad and shallow valleys. EQ'ing dips brings out distortion peaks at where you fill the dip. Generally I only chop peaks and smooth broad humps. In the end, the closer the headphone is already to the ideal, the better it will be. Like my grandmother said, "you can't make chicken pie out of chicken sh_t."